Difference-in-Differences

Question 1

collection of panel/longitudinal data

Answer 1

allows for estimation of baseline differences and understanding how treatment effects evolve over time

helps account for fixed differences

Question 2

what does DD do?

Answer 2

takes the difference between equations (at t=0 and t=1) to try and eliminate OVB and deliver the true effect

Question 3

parallel trends assumption

Answer 3

gap/differences between groups will be similar over time so it’s fine if they’re different

can’t be diverging or converging

Question 4

when is OVB not a problem?

Answer 4

when OVs are not correlated with the explanatory variable of interest
- randomisation

if we have panel data and the extent of OVB is constant over time, then a DD estimate delivers the true effect

Question 5

when is DD a sound approach?

Answer 5

when unobserved factors that could affect the outcome are stable over the study period